Sources

The data was sourced from three places:

The point-by-point data that was crucial to this project was collected by Jeff Sackmann and is posted on a github repository: https://github.com/JeffSackmann/tennis_slam_pointbypoint
The historical ELO ranking data was collected from the Ultimate Tennis Statistics Website: https://www.ultimatetennisstatistics.com/rankingsTable

Code Index

eda.ipynb: Just a data exploration, I’m keeping it there for my own purposes
Point_probs.ipynb: This is the notebook where I make the dataset and train the logistic model for predicted probabilities of success, the code saves the model as ‘point_prob_model.sav’ for use in streaks.py
Streaks.py: This is the script where the majority of the simulation and calculation goes on. The results of the simulation are stored in streaks_points.csv, which are analyzed in results_analysis.ipynb
Results_analysis.ipynb: This is where the results are analyzed and plots produced. Two of them are nicely interactive.

Name		Name	Last commit message	Last commit date
Latest commit History 14 Commits
plots		plots
tennis_data		tennis_data
Elo_Rankings2017.csv		Elo_Rankings2017.csv
Isner_model_check.png		Isner_model_check.png
Project_Guidelines.pdf		Project_Guidelines.pdf
README.md		README.md
eda.ipynb		eda.ipynb
federer_data.sav		federer_data.sav
federer_test_data.csv		federer_test_data.csv
point_prob_model.sav		point_prob_model.sav
point_probs.csv		point_probs.csv
point_probs.ipynb		point_probs.ipynb
results_analysis.ipynb		results_analysis.ipynb
streaks.py		streaks.py
streaks_points.csv		streaks_points.csv